Abstract
Since the introduction of Fourier transform NMR spectroscopy, free induction decay (FID) has been available in a binary format, which is required to apply several mathematical treatments on the raw data. Those mathematical procedures are the application of window functions on the FID, baseline correction of the time domain data, and finally the Fourier transformation. Nevertheless, it took a long time until digital filtering was applied on NMR raw data. Digital filtering can be done as either a postprocessing procedure or on the fly. The on the fly concept is preferable, but it requires specialized and very fast processors. With the common availability of such processors, called digital signal processors (DSP), digital filtering on NMR raw data became a routine method. The application of digital filters allows us to overcome problems such as baseline distortion in the NMR spectrum due to nonperfect phase response of any common analog antialiasing filter, distortion of the signal amplitude due to a nonperfect amplitude response of any common analog antialiasing filter, a limited dynamic range due to aliased noise by nonoversampling techniques, reduced dynamic range due to quantization noise, and folded signals due to folded signals outside of the spectral window. The manner in which digital filters are used in a NMR spectrometer is the same as audio applications; it just requires stronger processors to apply highβperformance digital filters on the fly. This article is an introduction to the concept of digital filters. Therefore, the description is strongly simplified and the ideal case is assumed. Β© 2002 Wiley Periodicals, Inc. Concepts in Magnetic Resonance (Magn Reson Engineering) 15: 164β176, 2002